Deep Event Stereo Leveraged by Event-to-Image Translation

Depth estimation in real-world applications requires precise responses to fast motion and challenging lighting conditions. Event cameras use bio-inspired event-driven sensors that provide instantaneous asynchronous information of pixel-level log intensity changes, which makes them suitable for depth such However, as the event primarily spatially sparse data, it is hard accurate dense disparity map stereo camera setups - especially estimating disparities on local structures or edges. In this study, we develop a novel deep network reconstructs spatial image features from embedded streams leverages using reconstructed compute maps. To end, propose event-to-image translation with cross-semantic attention mechanism calculates global semantic context reconstruction. addition, feature aggregation module developed estimation, modulates by stacked dilated spatially-adaptive denormalization mechanism. Experimental results reveal our method can outperform state-of-the-art methods significant margins both quantitative qualitative measures.